WITH weak GDP growth and falling real wages, the government does not have much to shout about on the economic front. Ministers, however, have one choice boast. It may defy popular belief but Theresa May, the prime minister, says that “inequality has been reduced to its lowest level in 30 years.” Philip Hammond, the chancellor, likes to wheel out this figure whenever he can. He may do so again on March 13th when he delivers his spring statement, a half-yearly update on the economy, to Parliament. But there is one problem with the government’s claim: it may not be true.

It is not that ministers are lying or massaging the figures. They appear to be referring to a release from the respected Office for National Statistics. The ONS’s figures, in turn, have received the seal of approval from the official statistics watchdog.

The release reports the Gini coefficient, where 1 represents complete income inequality and 0 represents complete equality. Unlike some other measures of inequality, such as the ratio between incomes at the 90th and 10th percentile, the Gini takes in the entire income distribution. Many economists therefore like it. The ONS’s measure suggests that in 2015-16 the Gini coefficient for disposable household income (that is, after tax is subtracted and welfare payments added), was 0.32. That is far higher than in the late 1970s, when the series begins. But it is indeed the lowest since 1986, as Mrs May claims.

But economists question the ONS’s figures. Though the government collects a huge amount of data on incomes, taxes and benefits, most related statistics are calculated with the help of surveys. For its inequality statistics, the ONS uses an annual survey of about 5,000 households, in which, among other tasks, every participant aged over 15 keeps a detailed record of their spending for two weeks.

Surveys, though, are far from perfect—and they are particularly bad at capturing the very rich, who matter hugely to inequality, since they are responsible for a big share of overall incomes. Tax data suggest that the top 1% of earners rake in up to a tenth of total household income, a higher proportion than in previous decades.

The ONS’s survey, small as it is, will cover perhaps 50 households in the top 1%. What is more, plutocrats may not report their incomes accurately in the survey. Rich people appear more likely to downplay their earnings. Research suggests that the top 1% under-report their incomes by 10-20%. The upshot is that the ONS’s measure of inequality may be inaccurate. (The ONS plans to tackle the problem in part by raising the number of people surveyed.)

What does this mean for the government’s claim that inequality is at a 30-year low? An alternative measure suggests that it is wrong. Wonks at the Department for Work and Pensions estimate the incomes of the very richest by supplementing their survey with data from tax returns. Because of the adjustment to top-income data, the DWP’s figures are the gold standard for measuring inequality, according to Adam Corlett of the Resolution Foundation, a think-tank. They show that inequality is higher than the ONS believes (see chart). And they contradict the government’s claim: in line with popular perception, income inequality has been creeping up.